AI pharma is hitting a phase II clinical success rate bottleneck: while AI can efficiently optimize preclinical molecules, its ability to predict human efficacy remains insufficient, and the phase II success rate of AI-discovered drugs is on par with that of traditional drugs—prompting the industry to shift toward human disease simulation. San Francisco-based AI digital twin company Differentia Bio has published a rare chain of prospective validation evidence: from 2023 to 2024, leveraging multi-omics data, AI knowledge graphs, and quantitative disease mechanism models, it predicted potential intervention directions such as CD19+ and IL-4/Th2 for IgG4-related disease and made its academic records public; in 2025, a CD19+-targeted drug received FDA approval, and in 2026, the IL-4/Th2 pathway garnered support from independent clinical research. This pattern—prediction first, independent validation later—forms a competitive barrier that is difficult to replicate. The company boasts a top-tier scientific team spanning multiple disciplines and relies on its self-developed DHARMA™ AI disease digital twin engine, integrating multiple technology types to build a multi-scale human disease simulation system. Unlike the mainstream approach that starts from the molecular end, it begins with the disease biology of real patients, enabling generative AI and computable human disease models to work in concert and form a closed loop of R&D decision-making—creating value in three areas: preclinical candidate drug screening, clinical development trial design, and milestone R&D decision-making. Global regulation is entering a policy window for model-informed drug development, and its digital twin models meet regulatory requirements for model credibility and other aspects. The company has already entered into commercial collaborations with multiple biopharmaceutical organizations and achieved profitability through simulation revenue. Its human disease simulation approach is driving drug R&D toward a new paradigm of virtual rehearsal coupled with real-world clinical validation, and is expected to systematically reduce clinical failures and improve R&D efficiency.